Change Log

0.10.0 - 2019-10-01

• Add support for HeapProfSampleEnd and HeapBioProfSampleBegin (#52)

0.9.1 - 2019-09-03

• Relax upper version bounds to support GHC 8.8.1 (#49 and #51)

0.9.0 - 2019-05-15

• Support the newly added par_balanced_copied field (#47)

0.8.0.2 - 2019-04-02

• Tighten lower version bound for base (#46)

0.8.0.1 - 2018-10-22

• Relax upper version bound for base to support GHC 8.6
• Relax upper version bound for binary

0.8.0 - 2018-07-11

• Add HeapProfBreakdownClosureType (#33, #39)
  • This is a breaking change
• Test with newer GHCs (#40)

0.7.3 - 2018-07-10

• Fixed memory-leak in incremental readEvents (#37)
• Relax upper version bound for containers (#38)

0.7.2 - 2018-03-13

• Add Semigroup instance for MaxVars to build with ghc-8.4

0.7.1 - 2018-02-17

• Export HeapProfBreakdown, HeapProfFlags, and PID types (#35)

0.7.0 - 2017-10-04

• Add support for heap profiling events (#29)

0.6.0 - 2017-05-31

This contains breaking changes.

• The deprecation notice on readEventLogFromFile has been retracted
• The incremental API has been refactored

The details are as follows:

• Update bug tracker URL (#10)
• New test for Eden events (#11)
• Relax version bound for binary (#15)
• Enable Travis CI (#19)
• Refactor the incremental API which was introduced in 0.5.0.0 (#22)
• Some speed/memory usage improvements (#18, #22)

0.5.0.0 - unreleased

• Readme added :)
• Old parser replaced with an incremental implementation
• General overhaul of the codebase
• Partial Haddock coverage

The 0.5.* releases should be able to handle large event logs and logs that have been cut off abruptly, e.g. from executable runs that resulted in crashes.

This release should be mostly backwards compatible, however the "old" way of reading event logs, namely the readEventLogFromFile function is now deprecated.

NOTE: Users parsing large logs may notice that this version of the library is noticably slower than the older versions. The incremental parser is at fault here - previous versions of the libray used a "trick" that would let them essentially skip the first step of the mergesort algorithm since EventBlocks were already sorted in time order. The new parser reads the file incrementally and cannot take the advantage of this. Apologies for any inconvenience this may have caused.